ML Research Wiki / Benchmarks / Unsupervised Domain Adaptation / ImageNet-C

ImageNet-C

Unsupervised Domain Adaptation Benchmark

Performance Over Time

📊 Showing 16 results | 📏 Metric: mean Corruption Error (mCE)

Top Performing Models

Rank Model Paper mean Corruption Error (mCE) Date Code
1 ResNet50 (baseline), BatchNorm Adaptation, 8 samples Improving robustness against common corruptions by covariate shift adaptation 65.00 2020-06-30 📦 bethgelab/robustness 📦 Claydon-Wang/OFTTA
2 ResNet50 (baseline), BatchNorm Adaptation, full adaptation Improving robustness against common corruptions by covariate shift adaptation 62.20 2020-06-30 📦 bethgelab/robustness 📦 Claydon-Wang/OFTTA
3 ResNet50 + ENT If your data distribution shifts, use self-learning 51.60 2021-04-27 📦 bethgelab/robustness
4 ResNet50 + RPL If your data distribution shifts, use self-learning 50.50 2021-04-27 📦 bethgelab/robustness
5 ResNet50+DeepAug+AugMix, BatchNorm Adaptation, 8 samples Improving robustness against common corruptions by covariate shift adaptation 48.40 2020-06-30 📦 bethgelab/robustness 📦 Claydon-Wang/OFTTA
6 ResNet50+DeepAug+AugMix, BatchNorm Adaptation, full adaptation Improving robustness against common corruptions by covariate shift adaptation 45.40 2020-06-30 📦 bethgelab/robustness 📦 Claydon-Wang/OFTTA
7 ResNeXt101 32x8d + ENT If your data distribution shifts, use self-learning 44.30 2021-04-27 📦 bethgelab/robustness
8 ResNeXt101 32x8d + RPL If your data distribution shifts, use self-learning 43.20 2021-04-27 📦 bethgelab/robustness
9 ResNeXt101 32x8d + IG-3.5B + RPL 📚 If your data distribution shifts, use self-learning 40.90 2021-04-27 📦 bethgelab/robustness
10 ResNeXt101 32x8d + IG-3.5B + ENT 📚 If your data distribution shifts, use self-learning 40.80 2021-04-27 📦 bethgelab/robustness

All Papers (16)

Improving robustness against common corruptions by covariate shift adaptation

2020
ResNet50 (baseline), BatchNorm Adaptation, 8 samples
bethgelab/robustness Claydon-Wang/OFTTA

Improving robustness against common corruptions by covariate shift adaptation

2020
ResNet50 (baseline), BatchNorm Adaptation, full adaptation
bethgelab/robustness Claydon-Wang/OFTTA
bethgelab/robustness
bethgelab/robustness

Improving robustness against common corruptions by covariate shift adaptation

2020
ResNet50+DeepAug+AugMix, BatchNorm Adaptation, 8 samples
bethgelab/robustness Claydon-Wang/OFTTA

Improving robustness against common corruptions by covariate shift adaptation

2020
ResNet50+DeepAug+AugMix, BatchNorm Adaptation, full adaptation
bethgelab/robustness Claydon-Wang/OFTTA
bethgelab/robustness
bethgelab/robustness

If your data distribution shifts, use self-learning

2021
ResNeXt101 32x8d + IG-3.5B + RPL
bethgelab/robustness

If your data distribution shifts, use self-learning

2021
ResNeXt101 32x8d + IG-3.5B + ENT
bethgelab/robustness

Improving robustness against common corruptions by covariate shift adaptation

2020
ResNeXt101+DeepAug+AugMix, BatchNorm Adaptation, 8 samples
bethgelab/robustness Claydon-Wang/OFTTA

Improving robustness against common corruptions by covariate shift adaptation

2020
ResNeXt101+DeepAug+AugMix, BatchNorm Adaptation, full adaptation
bethgelab/robustness Claydon-Wang/OFTTA

If your data distribution shifts, use self-learning

2021
ResNeXt101 32x8d + DeepAug + Augmix + ENT
bethgelab/robustness

If your data distribution shifts, use self-learning

2021
ResNeXt101 32x8d + DeepAug + Augmix + RPL
bethgelab/robustness
bethgelab/robustness
bethgelab/robustness